liveBook · Manning

Child document

As explained in Essential GraphRAG

A child document is a smaller text segment derived from a parent document and associated with an embedding for retrieval.

A child document is a smaller text segment derived from a parent document and stored with an embedding for retrieval. Splitting at structural boundaries first preserves coherent ideas before producing child chunks. section 3.2

A parent-document retrieval pipeline typically:

  • Splits a PDF into parent documents of up to 2,000 characters. section 3.2

Retrieving context from a paper

For “Einstein’s Patents and Inventions,” section-based splitting produces parent chunks, each containing 500-character child chunks. Query matching uses child embeddings, while retrieval can return the larger parent for richer context. section 3.2

How Essential GraphRAG covers it

How deep it goes
Explains structural splitting, chunk sizes, embeddings, graph storage, and vector indexing, but not retrieval-query execution.
Examples
Uses a research paper, section splitting, 2,000-character parents, 500-character children, and a Neo4j graph.
What it assumes
Assumes familiarity with document chunking, embeddings, Cypher, and graph database operations.
What you can do after
Builds a parent–child retrieval graph with embedded child nodes and a vector index.
How it connects
Connects child documents to parent documents, embeddings, vector indexes, PDFs, and graph databases.
Cover of Essential GraphRAG
liveBookmanning.com
Read Essential GraphRAG on liveBook Get the book at manning.com

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Essential GraphRAG. The section links point to the parts of the book it draws on.